光学学报, 2009, 29 (s1): 266, 网络出版: 2009-06-25
人工神经网络对爆炸物太赫兹光谱的识别
Identification of Terahertz Absorption Spectra of Explosives by Artificial Neural Networks
光谱学 太赫兹光谱 爆炸物识别 自组织神经网络 多层感知器 spectroscopy terahertz spectra identification of explosives self-organized mapping muti-layer perceptron
摘要
为了研究爆炸物在太赫兹波段的光谱特性, 进而对爆炸物进行检测和识别, 利用太赫兹时域光谱系统(THz-TDS),对RDX, HMX,DNT,PETN,TNPG 五种纯品炸药以及以RDX为基底的混合炸药(8701,PW0,R791,R852,塑性炸药(SU-1))和以HMX为基底的混合炸药(8702)的太赫兹吸收光谱在真空条件下进行了测量。然后用两种人工神经网络(ANNs)—自组织(SOM)神经网络和多层感知器 (MLP)神经网络—对爆炸物吸收光谱进行了识别, 经过不断地学习和训练, 取得了较好的鉴别结果, 正确率高于95%。实验结果表明, 用两种神经网络可以实现对纯品炸药和混合炸药的识别, 为太赫兹光谱技术用于爆炸物的检测和识别提供了一种有效的方法。
Abstract
In order to study the spectroscopic characters of explosives in the terahertz region and realize the identification of explosives, the terahertz absorption spectra of explosives have been measured by using terahertz time-domain spectroscopy (THz-TDS) system. These samples which include pure explosives RDX, HMX,DNT,PETN, TNPG and mixed explosives 8701,PW0,R791,R852,SU-1,8702 were then identified by two types of artificial neural networks(ANNs)—self-organized mapping(SOM)and muti-layer perceptron(MLP)—through repetitive modeling and adequate training. High positive identification rate (above 95%) and low false alarm rates have been gained. The results indicate that it is feasible to apply these two ANNs on the identification of different types of explosives, and it also provides an effective method in the inspection and identification for explosives using THz-TDS.
李微微, 冯瑞姝, 周庆莉, 张存林. 人工神经网络对爆炸物太赫兹光谱的识别[J]. 光学学报, 2009, 29(s1): 266. Li Weiwei, Feng Ruishu, Zhou Qingli, Zhang Cunlin. Identification of Terahertz Absorption Spectra of Explosives by Artificial Neural Networks[J]. Acta Optica Sinica, 2009, 29(s1): 266.